DeepCell Kiosk: Scaling deep learning-enabled cellular image analysis with Kubernetes

Abstract

Deep learning is transforming the ability of life scientists to extract information from images. These techniques have better accuracy than conventional approaches and enable previously impossible analyses. As the capability of deep learning methods expands, they are increasingly being applied to large imaging datasets. The computational demands of deep learning present a significant barrier to large-scale image analysis. To meet this challenge, we have developed DeepCell 2.0, a platform for deploying deep learning models on large imaging datasets (>10 5megapixel images) in the cloud. This software enables the turnkey deployment of a Kubernetes cluster on all commonly used operating systems. By using a microservice architecture, our platform matches computational operations with their hardware requirements to reduce operating costs. Further, it scales computational resources to meet demand, drastically reducing the time necessary for analysis of large datasets. A thorough analysis of costs demonstrates that cloud computing is economically competitive for this application. By treating hardware infrastructure as software, this work foreshadows a new generation of software packages for biology in which computational resources are a dynamically allocated resource.

Classifying T cell activity in autofluorescence intensity images with convolutional neural networks

“…Others do not contain any T cells. Developing a better filter to detect images with artifacts and adopting state-of-the-art segmentation approaches [16,50] could further boost classification accuracy. We highlight the images that are misclassified by the pre-trained CNN with fine-tuning on the UMAP plots for three alternative image representations (Figures S12-S14).…”

“…Advanced machine learning models, in particular convolutional neural networks (CNNs), are now the prevailing approach for a variety of cellular image analyses [10][11][12]. CNNs can classify cell phenotypes [13,14], segment cells [15,16], restore images [17], and predict protein localization [18,19], cell lineage choice [20], the biological activity of small molecules [21] or ratios of activated T cells in a population [22]. They are also effective at cell type classification tasks such as predicting cell cycle state [23] and cell sorting [24].…”

Multiplexed Single-cell Metabolic Profiles Organize the Spectrum of Cytotoxic Human T Cells

“…Background (due to absence of tissue or high gold signal) was removed, and noise was filtered out using a k-nearest-neighbor approach. To segment single cells from images, we trained a convolutional neural network 76,117 using annotated training data from a variety of different cancer types. The network output was fed into the watershed algorithm to produce individual cells.…”

BactMAP: An R package for integrating, analyzing and visualizing bacterial microscopy data

“…To make BactMAP as useful as possible for the bacterial cell biology community, we standardized the import of datasets from five popular segmentation programs and one single-molecule tracking program: SuperSegger (Stylianidou et al, 2016), MicrobeJ (Ducret et al, 2016), Oufti (Paintdakhi et al, 2016), Morphometrics (Ursell et al, 2017), ObjectJ/ChainTracer (Syvertsson, Vischer, Gao, & Hamoen, 2016;Vischer et al, 2015) and iSBatch (Caldas, Punter, Ghodke, Robinson, & Oijen, 2015). Furthermore, new programs with improved or specialized cell segmentation capabilities are being developed constantly, as recently for instance BacStalk (Hartmann, van Teeseling, Thanbichler, & Drescher, 2018), a specialized tool for segmentation of cells with complex morphologies and the deep learning and machine-learningbased segmentation programs DeepCell (Bannon et al, 2018), DeLTA (Lugagne, Lin, & Dunlop, 2019) and Ilastik (Berg et al, 2019). Therefore, to make BactMAP compatible with a wide range of current and future segmentation software packages, we also implemented a generic import function for segmentation input.…”